Unreliability of β-2-microglobulin in early detection of central nervous system relapse in acute lymphoblastic leukemia

0277.5379~85$3.00+0.00 1985 Prrpmon Pw5s Ltd.

0

Unreliability of P-Z-Microglobulin in Early Detection of Central Nervous System Relapse in Acute Lymphoblastic Leukemia AD F. NAGELKERKE,*t

GERARD

of Pediatrics

*Department

University

Abstract-The

value

microglobulzn

(f&m)

(ALL)

in the central

children.

Hospital,

level for nervous

system

(95% confidence proved

significantly

to

be

conditions,

inuolrxment

assay. Results

5.3-85.3%)

than

of white

in Incipient

are more

was found

often

between

in children

with acute

lymphoblustic

syndrome,

lymphoblastic

IN ACUTE

system (CNS)

CNS

associated

therapeutic

problems.

(ALL)

Screening

protein

investigation

level.

might

Early

improve

years reports ment

CSF

were elevated findings

detection

for

[2]. Mavligit

of CNS

relapse

ALL.

studies

with

of&m

serial

1982 from all children of

during

The

Study Group. determination routine protocol

and

in serum and

clinical

June

the

at the Depart-

Free

University

were treated according

of the Dutch

at

1981 to

with ALL (n = 30)

puncture of

patients

were collected

the period

lumbar

Pediatrics

the protocols

ratio

that

concluded

August

Hospital.

[l] stated

Consequently,

intervals

ment

these

involvement in or lymphoma.

regular

who underwent

(&m)

support

AND METHODS

Blood serum and CSF samples

with CNS involve-

and co-workers

determination

in childhood

Patients

In the past

that Pnm levels in CSF and the CSF/serum

simultaneous

as a test for early

of CNS involvement

MATERIALS

of

were well correlated with CNS patients with acute leukemia they

determination

CNS

of beta-Z-microglobulin Recent

of CNS

of &m

detection

central

that in patients

or lymphoma [l].

these

and&m

test for early detection

sampling of count, cyto-

the results of therapy.

levels

and

leukemia.

and determination

have indicated

acute leukemia

relapse,

is one of the major

leukemia is performed by periodical cerebrospinal fluid (CSF) for cell pathological

as well as in

with Bnm elevation

value

leukemia relapse

a

value of 8%

cells in CSF, Bnm values

INTRODUCTION nervous

Free

the Netherlands

in 30

demonstrated

CSF white cell counts

that Bnm is not an appropriate

of Immunochemtstry,

leukemia

studied

and a predictive

In post-irradiation

higher

Division

fluid (CSF) B-Z-

lymphoblastic

has been prospectively

changes

cause,

No relation

levels. It is concluded

interval,

and FRANS cl. DE WAAL*

J. P. VEERMAN* Chemistry,

1117, Post Box 7057, 1007 MB Amsterdam,

of acute

by micro-ELISA

cytopathological

or an unknown

than CNS relapse.

(CNS)

l.O-26.0%).

interval, with

ANJO

of Clinical

of the cerebrospinal

early detection

of 40% (95% confidence

viral infection

KAMP,$

De Boelelaan

of serial determination

Bsm was determined

sensitivity

J. VAN

and IDepartment

Childhood

The frequency depended on the

to

Leukemia

of CSF-&rn frequency of

sampling according to the treatment (every 14 weeks), and on the presence of indications

to perform a lumbar puncture.

CSF might be useful in the early diagnosis of CNS involvement and in the monitoring of intrathecal

CSF was obtained by lumbar patients), or from an Ommaya

therapy in patients with acute leukemia lymphoma. The aim of the present study is to investigate

patients). The minimal follow-up period was 12 months. Diagnosis of CNS-leukemia was based on the

Acteptcd tTo

or the

puncture (26 reservoir (four

presence of malignant lymphoblasts in cytopathological examination of two samples of CSF taken with an interval of 2-7 days.

12 De<-ember 1984.

whom rrqu~rta for reprints should be addressed.

659

660

A. F. Nagelkerke

Methods P2rn was determined by a micro-ELISA assay. The test as used in our study is the test described by Ferrua et al. [3], using commercially available antibodies. The wells of the microtitration plate were coated with 200 ~1 of diluted &m antibodies and left overnight at 37°C. The plate was emptied by inversion and washed three times with 220 ~1 wash buffer in each well. Between washes the buffer was left for at least 5 min in the plate, reducing non-specific adsorption. The wells were filled with 100 ~1 of &m standard or samples, if necessary diluted in phosphate-buffered saline (PBS) containing 0.3% bovine serum albumin (BSA), followed by incubation for 75 min at 37°C. After another washing step, 190 ,ul substrate solution was added and the plate was incubated at room temperature, in the dark, for 30 min. The enzyme reaction was stopped by the addition of 45 ~1 of 4 N HCl and the intensity of the color developed was read at 492 nm. Materials Rabbit &m antibodies (DAKO-Immunochemicals, Copenhagen, Denmark, code A 072), diluted l/2000 with coating buffer (0.1 M NaHCO,, pH 9.6). Solid phase: 96-well flat bottom microtiter plate, Nunc, No. 4-42404. flnrn standard was from the Phadebas &mmicrotest (kindly donated by Pharmacia Diagnostics, Uppsala, Sweden). Rabbit &m-antibodies were conjugated with radish peroxidase (DAKO-Immunohorse chemicals, code P 174) and diluted 112000 in PBS (pH 7.4: 1.22 g K,HPO,, 8.77 g NaCl in 800 ml distilled water, adjusted to pH 7.4 with 4 M NaOH and made up to 1 1). The wash buffer was 0.3% (v/v) BSA (Poviet, Organon Teknika. Oss, Holland) and 0.1% (v/v) Tween-20 in PBS, The substrate solution was made up of 20 mg 1,2-phenylene diamine (OPD) and 0.15 ml 30% H,O, in 20 ml citric acid-phosphate buffer(0.1 M citric acid adjusted to pH 5.3 with 0.2 M Na2HP04). The intensity of the color developed after incubation was measured with an automatic micro-ELISA reader system (Organon Teknika, Oss, Holland). The results obtained for serum and urine samples of healthy control persons compare well with those described by Ferruaetal. [3]. Moreover, a good correlation was found comparing our test with the Behringwerke test for determination of human &m (Enzygnost). In 12 sera a correlation coefficient of r = 0.95 and in 18 urines of r = 1.OO was found. The concentration of serum Pnm in

et al.

healthy control subjects, as measured with our test system, all fell within the normal distribution as described by Teasdale et al. [4], analysing the sera with the Pharmacia Phadebas [1211]/3-2-microglobulin test. The within-assay variation of our test was 2% (n = 20) and the day-to-day variation was 7.5% (n = 16). Normal values Normal values of CSF-Psrn reported in previous studies are 1150 f 370 ng/ml (mean f S.D.) and 1100 f 500 ng/ml [5, 61; ranges have an upper limit of 2000 ng/ml [2, 71. Reported levels in patients with leukemia and lymphoma without CNS involvement are 2400 f 300 ng/ml and 1400 f 660 ng/ml [ 1,2]; Demeocq mentions a lower level: 980 f 460 ng/ml in patients without or with only minor CNS involvement [B]. Regarding the CSF/serum ratio, Tenhunen et al. reporteda levelof 0.79 f 0.32 [5]; Mavligitet al. used a limit of limit of 1.0 [l]. In the absence of established limits for the normal range, we selected criteria for the elevation of CSF-&m. As suggested by Mavligit et al. [ 11, we took in to account the CSF level itself, the trend in it and the CSF/serum ratio (Table 1). RESULTS One hundred and twenty-four paired (blood and CSF) samples were obtained and 30 single samples. The median lumbar CSF-&m level was 1050 ng/ml (range, 300-4850 ng/ml); the median level in ventricular fluid was 520 ng/ml (range, 115-2100 ng/ml). The median serum &m level was 1450 ng/ml (range, 610-4000 ng/ml). Sensitivity During the investigation period five patients showed a new CNS involvement. Data of these patients are listed in Table 2. In three of the patients with new CNS involvement CSF-&m was not elevated according to our criteria (patients, A, B, and C). One other patient (D) had a slight elevation, according to just one criterion (CSF/serum ratio 1.06). The CSF-P,m value in this patient was rather low (750 ng/ml); there was no trend to elevation either. The fifth patient (E) had a clearly elevated CSF-&m. This patient, however, appeared to Table 1. (a) CSF-&m

Criteria for eleuation of CSF-P,rn

2 1500 ng/ml

(b) Elevation of CSF-&m 2 500 ng/ml compared with the previous determination (c)

Decline of CSF-&m 3 500 ng/ml at the following determination

(d) CSF/serum

ratio of &m > 1

and/or and/or and/or

CSF P-2-Microglobulin Table 2.

&m levels in CSF in patients

in CNS-ALL

with new, untreated

Patient A* Bi CS

420 1100 460

425 (7) 700 (14)

Dt

750

750 (14)

E§

-1350 II

-

have contracted a viral meningitis 2 months earlier, which might have been the cause of the &m elevation [7]. Thus in this small patient group the sensitivity did not exceed 40% (95% confidence interval, 5.3-85.3%). value

of positiue

test

All cases of elevated CSF-&m elevation are listed in Table 3. It is clear that CSF-&m elevation is associated with new CNS involvement only in a small minority of cases. In this patient group the predictive value of a positive test is 8% (2/25; the four patients already treated for previously diagnosed CNS involvement were left out). The 95% confidence interval is l.O-26.0%. Table 3 shows that in our patients &rn elevation was associated with various conditions. The most important of these were postirradiation syndrome and viral infection. Post-irradiation syndrome. During the investigation period six patients received a 25-Gy CNS irradiation within 2-3 weeks as a part of CNS prophylaxis according to the protocol. In all of these patients the CSF-&.m level was elevated during the period from 9 to 19 weeksafter the start of the irradiation. During this period in all cases PPrn was higher than the last level measured before and the first level measured after this Table

3.

Numbers

of patients &m

Post-irradiation Viral changes CNS

with

450 (3) 1100 (2) 820 (2)

CNS

Cerebral Total

6

treatment

4

chemoprophylaxis

untreated

After a general Cause

6

cytopathological

during CNS

toxoplasmosis unknowm

leukemia

infection

0.29 0.81 0.44

540 (1%)

1.0611

575 (1)

0.48

ratio

> 1; patient

E:

period. The difference is statistically significant (signed rank test, P < 0.05). All but one of these patients showed the clinical the post-irradiation features of syndrome (somnolence, malaise, fever) during the same period. Viral infection. Six patients showed cytopathological changes in CSF cells during a minor general or respiratory infection of apparently viral origin. Clinically there were no signs of meningeal irritation. Four of these patients had a mild pleocytosis consisting of mononuclear cells (range, 6-31/mm3). In cytological examination these cells appeared to be for the most part activated lymphocytes, but sometimes resembled malignant lymphoblasts. All six patients had elevated CSFP2rn, which returned to normal without specific treatment within a few weeks, together with the disappearance of pleocytosis and abnormal mononuclear cells. During a follow-up period of 1 yr none of these patients developed a CNS-relapse of ALL. Further

statistical

analysis

CSF levels of j3*rn and CSF/serum ratios in the patient groups with post-irradiation syndrome, viral infection with cytopathological changes of

CSF-

in CSF leukemia,

During New,

with an elevated

CSFiserum ratio

level

syndrome

infection

CNS inuolvement

&n in CSF (n,q:ml) Previous level Next level (interval in weeks) (interval in weeks)

At time of CNS relapse

*Isolated CNS relapse; CSF from Ommaya reservoir. TIsolated CNS relapse. SRelapse in bone marrow and CNS simultaneously. §Initial ALL in bone marrow and CNS; after viral meningitis. lIThe elevated &m levels are underlined. [Patient D: CSF/serum decline > 500 ng/ml at the next determination (see Table l).]

Predictive

661

Table patient

4.

cells (VIR)

ration

and with untreated

_

(mean

of &rn in

syndrome

with cytopathological (CNS-ALL)

2 1

changes

isolated f

PIRS VIR CNS-ALL

(ng/ml)

2825 f 582.6 1840 f 878.6 682 f 314.8

(PIRS), of CSF-

CNS leukemia

S.D.) CSF/serum

CSF-P,m

1 29

with post-irradiation

viral infection

2

7

CSF level and CSFlserum

groups

ratio

of &m 2.06 + 0.220 1.19 f 0.076 0.67 f 0.366

A. F. Nagelkerke

662

CSF cells and untreated isolated CNS relapse of ALL are listed in Table 4. The &m level in CSF and the CSF/serum ratio proved to be significantly lower in untreated isolated CNS-ALL then in both post-irradiation syndrome and viral infection with CSF-cell changes (Wilcoxon’s two-sample test, P < 0.05). Furthermore, the CSF/serum ratio was significantly higher in the post-irradiation syndrome than in viral infection (P < 0.05). There was no evidence of a correlation between CSF white cell count and CSF-&rn level (correlation coefficient r = -0.021). Ommaya reservoir Ventricular fluid might have a lower level of psrn than CSF obtained by lumbar puncture. In one patient we simultaneously determined a ventricular Pnm level of 440 ng/ml and a lumbar level of 820 ng/ml. Of the five patients who developed a CNS relapse during the investigation period only one had an Ommaya reservoir at that time. In this patient the CSF-P,rn level showedno change at the moment of CNS relapse. Despite the absence of normal values of CSF-&m in ventricular fluid, we can assume that in this patient CSF-P,m could not have been helpful in detecting CNS relapse. DISCUSSION Methods

Our results seem to differ markedly from those of other authors [l, 21. Theoretically, the method of determination of &m could be responsible for that difference (micro-ELISA vs RIA assay). We think we have standardized and documented our determination method sufficiently to rule out this possibility. Until sufficient normal values are available the use of ventricular fluid for &m sampling should be avoided. Reference ranges established for lumbar CSF should not be applied to ventricular specimens. Consequently until now only the trend in CSF-&m can give information in these patients. As indicated before, this problem did not influence the results of this study. Normal

values

A test for detection of CNS leukemia should be a sensitive one. Hence the upper limits of normal chosen here are relatively low (Table 1). Choosing other limits (e.g. CSF-P,rn > 2000 ng/ml) would not have changed the results substantially, though. The use of the CSF/serum ratio, as indicated by Mavligit et al. [l], can be criticized because no evidence has been found of a relation between both levels [2, 71. If the CSF/serum ratio as a

et al.

criterion is abandoned the results of this study are only marginally influenced: in only 4/29 patients with elevation of &m (Table 3) was the elevation based merely on the CSF/serum ratio. Of these four patients one had a CNS relapse; the other three belonged to the group ‘cause unknown’. finrn and CNS

relapse

The present study is the first on this subject which is confined to children with ALL. Previous studies have described mixed patient groups [ 1,2,

81.

The results of this investigation demonstrate that serial determination of CSF-&m and/or the CSF/serum &m ratio cannot be regarded as useful tests for early detection of CNS involvement in childhood ALL, as was suggested by Mavligit and co-workers [ 11. The sensitivity of the test in this population of patients with incipient CNS involvement proved to be low. Previous reports did not record the sensitivity [l, 2, 81. Moreover, in these investigations the populations studied contained many patients with advanced stages of disease. In our opinion a prospective study, like the present one, in a population with no known CNS involvement is the best way to determine the utility of a test for early detection of CNS disease

PI.

Because of the wide confidence interval (5.3-85.3%) the sensitivity of CSF-/3,m in beginning CNS-ALL might eventually prove to be more favorable than the 40% found in this study. However, even if that proves to be true, the utility of the test remains low because of the low predictive value of a positive result (8%). This low predictive value is a striking finding. In the past years some patients have received intensive intrathecal chemotherapy based on the mere elevation of CSF-&m [l, lo]. Our results show that such a decision is not justified. Treatment should not be based on a test before the reliability of that test is evaluated in the population in which it is intended to use the test. The number of patients in this study is too low to draw conclusions regarding the value of Wrn as a parameter for the degree of CNS involvement or for monitoring the therapy. Post-irradiation

syndrome

The finding that all patients who received a CNS irradiation during the investigation period demonstrated a substantial rise in CSF-/3,m is in accordance with the reports of Demeocq et al. [8, 111, who have recordeda &rn elevation 4-10 weeks after termination of CNS prophylaxis including cranial irradiation.

CSF P-P-Microglobulin Viral injection

Six patients showed cytopathological changes in CSF cells during a viral infection. In this patient group the &m level in CSF was significantly higher than in untreated CNS leukemia. It is known that lymphocytes in CSF during a viral infection may mimic leukemic blast cells. This could lead to a false diagnosis of CNS leukemia [12], and might burden the patient with a long, heavy, dangerous and unnecessary treatment.

in CNS-ALL Because &m may be elevated both in viral infection and in CNS relapse it is not very helpful in distinguishing between both conditions.

Conclusion

We conclude that /I-2-microglobulin is not an appropriate test for the early detection of CNS involvement in children with acute lymphoblastic leukemia because of a low predictive value and a low sensitivity.

REFERENCES

2. 3.

4.

5. 6.

7. 8.

9.

10. 11.

12.

Mavligit GM, Stuckey SE, Cabanillas FF er al. Diagnosis of leukemia or lymphoma in the central nervous system by beta-2-microglobulin determination. N Engl JMed 1980, 303, 718-722. Koch Th R, Lichterfeld KM, Wiernik PH. Detection of central nervous system metastasis with cerebrospinal fluid beta-2-microglobulin. Cancer 1983, 52, 101-104. Ferrua B, Vincent C, Revillard JP et al. A sandwich method of enzyme immunoassay. III. Assay for human beta-2-microglobulin compared with radioimmunoassay. J Immunol Methods 1980, 36, 149-158. Teasdale C, Mander AM, Fifield R, Keyser JW, Newcombe RG, Hughes LE. Serum beta-2-microglobulin in controls and cancer patients, Clin Chim Acta 1977, 78, 135-143. Tenhunen R, Iivanainen M, Kovanen J. Cerebrospinal fluid beta-Z-microglobulin in neurological disorders. Acta Neurol Stand 1978, 57, 366-373. Gekle D, Kult J, Roth R. Lysozym und Beta-2-microglobulin in Liquor gesunder Kinder und bei Kindern mit Erkrankungen des Zentralnervensystemes. Klin Wochenschr 1977,55,189-191. Starmans JJP, Vos J, van der Helm HJ. The beta-2-microglobulin content of the cerebrospinal fluid in neurological disease. J Neural Sci 1977, 33, 45-49. Demeocq F, Malpuech G, Raynaud EJ, Plagne R, Gaillard G, Robert A. Beta-2microglobulin in cerebrospinal fluid of patients with acute leukemia of lymphoma. In: Raybaud C, Clement R, Lebreuil G, Bernard JL, eds. Pediatric Oncology (Proceedings of the XIIIth Meeting of the International Society of Pediatric Oncology). Amsterdam, Excerpta Medica, 1982, 355-356. Wulff HR. Rational Diagnosis and Treatment. Oxford, Blackwell, 1976, 90-96. Mavligit GM. Answer to letter to the editor. N Engl J Med 1981, 304, 362. Demeocq F, Malpuech G, Raynaud EJ, Plagne R, Gaillard G. Elevation of beta-2microglobulin in cerebrospinal fluid after irradiation of the central nervous system. N Engl J Med 1981, 304, 1366. Borowitz M, Bigner SM, Johnston WW. Diagnostic problems in the cytologic evaluation of cerebrospinal fluid for lymphoma and leukemia. Acta Cytol 1981, 25, 665-674.